Large disk storage systems like the 3380 and 3390 direct access storage devices (DASD) systems employed with many IBM mainframe computer systems are implemented utilizing many disk drives. These disk drives are specially made to implement a count, key, and data (CKD) record format on the disk drives. Disk drives utilizing the CKD format have a special "address mark" on each track signifying the beginning of a record on the track. After the address mark comes the three part record beginning with the "COUNT" which serves as the record ID and also indicates the lengths of both the optional key and the data portions of the record, followed by the optional "KEY" portion, which in turn is followed by the "DATA" portion of the record.
Although this format gives the system and user some flexibility and freedom in the usage of the disk drive, this flexibility forces the user to use more complicated computer programs for handling and searching data on the disk. Since the disk drive track has no physical position indicator, the disk drive controller has no idea of the data which is positioned under the read/write head at any given instant in time. Thus, before data can be read from or written to the disk drive, a search for the record must be performed by sequentially reading all the record ID's contained in the count field of all the records on a track until a match is found. In such a search, each record is sequentially searched until a matching ID is found. Even if cache memory is used, all the records to be searched must first be read into the cache before being searched. Since searching for the record takes much longer than actual data transfer, the disk storage system spends a tremendous amount of time searching for data which drastically reduces system performance.
Disk drives employing what is known as a Fixed Block Architecture (FBA) are widely available in small, high capacity packages. These drives, by virtue of their architecture, tend to be of higher performance than drives employing a CKD format. Such FBA drives are available, for example, from Fujitsu as 5.25" drives with 1 gigabyte or greater capacity.
The distinct advantage of utilizing many small disk drives is the ability to form a disk array. Thus a large storage capacity can be provided in a reduced amount of space, and storage redundancy can be provided in a cost effective manner. A serious problem arises, however, when trying to do a "simple" conversion of data from CKD formatted disks to FBA disks. Two schemes for such a conversion have been considered which do not provide an acceptable solution to the conversion problem. The first of such schemes involves placing every field i.e. Count, Key and Data, of the CKD formatted record into a separate block on the FBA disk drive. Although this scheme does not waste valuable disk space when CKD formatted records contain large amounts of data, the "Count" field which is very short (8 bytes) occupies an entire block which is typically at least 512 bytes. For example, a CKD formatted record containing 47K bytes of data could be converted to 95 blocks of FBA disk, 512 bytes in length. In such a conversion, one block would be used to store the count of the record while 94 blocks (47K bytes length of data divided 512 bytes of FBA disk block) would be used to store data, for a total of 95 blocks. However, search time for finding the desired record is still a problem since all the records must be sequentially searched.
For records having very short data lengths such as eight bytes, however, one full track, or 94 CKD formatted data records would need 188 blocks on the FBA disk: 94 blocks for the count portion of the records and 94 blocks for the data portion of the records, even though each data record may only occupy 8 bytes of a 512 byte FBA block. Such a scheme may thus waste nearly 50% of the disk space on an FBA disk drive.
The second scheme for converting data from CKD to FBA drives involves starting each CKD record in a separate block and then writing the complete record in sequential blocks. Utilizing such a scheme, the first FBA block will contain the "count" portion of the record as well as the optional key portion and the start of the data portion of the record. This scheme, however, produces serious system performance degradation when data must be written to the disk, since before writing data to the disk, the entire record must first be read into memory, modified, and subsequently written back to the disk drive. Such a loss in system performance is generally unacceptable.